Review Grok 4.5 token pricing, the 200K long-context threshold, cached-input rates, and additional server-side tool charges.
xAI prices Grok 4.5 by input, cached input, and output tokens. Requests whose prompt reaches 200K tokens use the higher long-context rate for every token in the request.
Example cost uses 1,000,000 input tokens and 300,000 output tokens at each model's default tracked rate, with no cache discount.
| Model | Input / 1M | Cached input / 1M | Output / 1M | Context | Example cost |
|---|---|---|---|---|---|
| Grok 4.5 Stable | $2.00 | $0.50 | $6.00 | 500K tokens | $3.80 |
Cost boundary: this example compares token charges only. It does not assume equal output quality, latency, retry rates, or output length across models.
Below 200K prompt tokens, the official rate is $2 per million input tokens, $0.50 per million cached input tokens, and $6 per million output tokens. Requests at or above the threshold use $4 input, $1 cached input, and $12 output per million tokens for the full request.
Yes. xAI documents separate per-call pricing for server-side tools such as Web Search, X Search, code execution, and file or collection search. StackLens token estimates do not silently include an assumed number of tool calls.
The official Grok 4.5 model page lists a 500,000-token context window. Its higher pricing tier begins when the prompt reaches 200,000 tokens.